JP7303504B2 - Image processing device, image processing method and program - Google Patents

Description

The present invention relates to an image processing apparatus, an image processing method, and a program for converting monochrome or color multi-valued input image data into multi-valued quantized output image data.

The error diffusion method is widely used as a binary quantization method for printed images. The error diffusion method propagates a quantization error during binary quantization to the surroundings, and reproduces a pseudo intermediate density by the density of black and white. The error diffusion method is simple in configuration and control, but has the problem that the spatial variation in binary quantization of the contour portion of the printed image increases, and the contour portion cannot be reproduced accurately.

To address this problem, the image processing apparatus, program, and method described in Japanese Patent Application Laid-Open No. 2002-200010 are designed to reduce the outline portion of printed characters that occurs when using the error diffusion method, which is generally used as a quantization method for printed images. Quantization that preserves the smoothness of edges by performing logical synthesis after performing quantization on edges and quantization on non-edges separately for the purpose of reducing spatial variations in quantization. is realized.

However, as shown in FIGS. 10 and 11 of an embodiment of Patent Document 1, Patent Document 1 targets characters having a clear luminance difference from the background, and for natural images such as photographs, Embodiment As shown in FIG. 5, only special portions with clear edges are processed, and basically it is devised so as not to affect photographs and the like. Therefore, it cannot be applied to the storage of textures in images, which is the object of the present invention.

In addition, although binary quantized image data is used in commercial printers, in the field of specialized printing such as newspapers and advertisements, binary quantized image data is not necessarily output, but rather to improve the expressive power of gradation. In addition, multilevel quantized images such as quaternary are often printed.

JP 2018-98552 A

The present invention realizes a technique for converting monochrome or color multi-valued input image data into multi-valued quantized output image data, and is applicable to enhancement of texture in images such as photographs, an image processing apparatus, an image processing method, and a program. intended to provide

In order to solve the above problems, an image processing apparatus according to the present invention is an image processing apparatus for converting multi-valued input image data into multi-valued quantized output image data. a level image generation unit for generating a plurality of different multivalued image data; a texture detection unit for detecting texture information of each of the multivalued image data;
an error diffusion unit for converting each multi-valued image data into each binary quantized image data by executing a gradation conversion process on each of said multi-valued image data so as to emphasize said texture information; and a multi-level quantized image generation unit that generates multi-level quantized output image data based on each binary quantized image data.

It is preferable that the texture information is surface unevenness information, surface pattern information, brightness information, edge information, or contour information.

It is preferred to have an error distribution unit for calculating the distribution of the quantization error of the transform according to the texture information.

It is preferable that the texture information is texture enhancement processing information of the multivalued input image data.

It is preferable that the texture enhancement processing information is frequency subband information.

a difference calculation unit that generates a difference image of each of the multilevel input image data and each of the binary quantized image data quantized by the error diffusion unit; and a visual processing unit that is a visual characteristic filter that inputs the difference image. a correction amount calculation unit for calculating a correction amount so as to preserve the texture information; and a multi-valued image correction unit for correcting each of the multi-valued input image data based on the correction amount. Preferably, the error diffusion section generates the binary quantized image data based on the output of the value image correction section.

An image processing method according to the present invention is an image processing method for converting multi-valued input image data into multi-valued quantized output image data, wherein the multi-valued input image data is generated by a level image generation section with a plurality of levels having different brightnesses. a texture detection step of detecting texture information of each multivalued image data by a texture detection unit; and each multivalued image data so as to emphasize the texture information. a step of converting each multi-valued image data into each binary quantized image data by an error diffusion unit by executing a gradation conversion process on the image data; and generating multi-level quantized output image data in a multi-level quantized image generation unit.

a step of generating a differential image of the binary quantized image data quantized by the error diffusion unit and each of the multi-valued input image data by a difference calculation unit; a step of calculating a correction amount by a correction amount calculation unit so as to store the texture information; and a step of correcting each of the multivalued input image data by the multivalued image correction unit based on the correction amount. and generating the binary quantized image data in the error diffusion section based on the output of the multi-valued image correction section.

It is preferable to repeatedly repeat the step of calculating the correction amount in the correction amount calculating section.

A program of the present invention is a program for causing a computer to function as each means of the image processing apparatus.

INDUSTRIAL APPLICABILITY According to the present invention, an image processing apparatus and an image processing method that realize a method of converting monochrome or color multi-valued input image data into multi-valued quantized output image data and that can be applied to enhance textures in images such as photographs. and a program can be provided. Furthermore, according to the present invention, by applying the processing of the present invention to images such as CMYK images obtained by separating color images, it is possible to process color multi-valued images.

1 is a block diagram showing a schematic configuration of an embodiment of an image processing device of the present invention; FIG. 4 is a flow chart showing the flow of one embodiment of the image processing method of the present invention; FIG. 3 is a flow chart showing the flow of a quantization process according to one embodiment in FIG. 2; FIG. FIG. 10 is a block diagram showing a schematic configuration of another embodiment of the image processing device of the present invention; 3 is a flow chart showing the flow of a quantization process according to another embodiment in FIG. 2; 1 is a schematic configuration diagram showing a printing apparatus using an image processing apparatus according to the present invention; FIG. It is an image showing the processing process (cyan component) of image processing using the image processing apparatus according to the present invention. 10A and 10B are images showing printing results of Example 1 and Comparative Example, where (a) is the result of Comparative Example 1 and (b) is the result of Example 1; 10A and 10B are images showing printing results of Example 2 and Comparative Example 2, where (a) is the result of Comparative Example 2 and (b) is the result of Example 2;

Embodiments of the present invention will be described below, but these embodiments are shown by way of example, and it goes without saying that various modifications are possible without departing from the technical idea of the present invention. Note that basically the same members are represented by the same reference numerals.

In the figure, reference numerals 10A and 10B denote respective embodiments of the image processing apparatus according to the present invention. The image processing apparatuses 10A and 10B are suitably used in, for example, a one-pass ink jet printer 100 as shown in FIG. In FIG. 6, the inkjet printer 100 prints C (cyan), M (magenta), Y (yellow), and K (black) on a web-shaped printing substrate 14 stretched over rolls 12a and 12b. It is equipped with inkjet heads 16a, 16b, 16c, and 16d for printing ink, and is controlled by the head controller 18 to eject ink from the inkjet heads 16a, 16b, 16c, and 16d.

The image processing apparatuses 10A and 10B output image data to the head control unit 18 so that the inkjet heads 16a, 16b, 16c, and 16d print on the web-like printing substrate 14. FIG.

<Image processing device 10A>
A schematic configuration of the image processing apparatus 10A is shown in the block diagram of FIG. As shown in FIG. 1, the image processing apparatus 10A is an image processing apparatus that converts multilevel input image data 20 into multilevel quantized output image data 42. For the multilevel input image data 20, brightness a level image generator 22 for generating a plurality of multi-valued image data 28 (multi-valued image data level 1 to multi-valued image data level N) having different values; and texture detection for detecting texture information of each multi-valued image data 28. A unit 30 performs tone conversion processing on each of the multi-valued image data 28 so as to emphasize the texture information, thereby transforming each of the multi-valued image data 28 into each of binary quantized image data 36 (binary image data). an error diffusion unit 34 for converting data 1 to binary image data N), and a multi-value quantized image generation unit 40 for generating multi-value quantized output image data based on each of the plurality of binary quantized image data 36. and an image processing apparatus. It also has an error distribution unit 32 having a function of determining the distribution of quantization errors in the error diffusion method according to the intensity of the texture information.

In the present invention, multilevel input image data 20 can be converted into N-bit multilevel quantized output image data 42 . Therefore, each of the plurality of multi-valued image data 28 has a number of levels from multi-valued image data level 1 to multi-valued image data level N according to the number of bits of the multi-valued quantized image to be output. Then, according to the number of each multivalued image data 28 (level 1 to level N), the quantization unit 26 converts each binary quantized image data 36 (binary image data 1 to binary image data N) into will be created.

The image processing apparatus 10A is a computer including a control unit such as a CPU and a storage unit, and is made to function as each means by a program so as to perform each process of the image processing apparatus 10A. Further, each process of the image processing apparatus 10A may be performed via a network such as the Internet.

The multi-valued input image data 20 is a color or monochrome digital image expressed in 8-bit gradation or the like, and is replaced with any non-binary image (for example, an image expressed in 8-bit format). quantization is performed to obtain multilevel quantized output image data 42 .

One embodiment of the image processing method by the image processing apparatus 10A will be described with reference to FIGS. 1 to 3. FIG. The image processing method can be executed by a program stored in a memory of a computer.

The image processing method of the present invention by the image processing apparatus 10A is an image processing method for converting monochrome or color multi-value input image data 20 into monochrome or color multi-value quantized output image data 42 . As an example, a case of outputting a four-level quantized image (2 bits: N=2) as the multi-level quantized output image data 42 will be described below with reference to FIGS.

First, monochrome or color multi-valued input image data 20 is input from another information terminal or storage medium (multi-valued image data input step, S100).

The level image generation unit 22 generates two types of multivalued images for the multivalued input image data 20 (multivalued image generation step, S102). At this time, any transform function may be used as long as two types of images, an image with enhanced bright portions and an image with enhanced dark portions, can be generated by nonlinear image conversion. for example,
Y=x ⁿ and Y=x ^1/n , 0≤n≤1
This can be achieved by setting two types of conversion functions. The quantization unit 26 performs two types of binary quantization on each of the multivalued image data 28 (multivalued image data level 1 and multivalued image data level 2), which are two types of generated images with different brightness. A quantized image is generated (quantization step, S106a).

The sub-steps in the quantization step S106a are described with reference to FIG.

When the two types of multivalued image data 28 generated by the level image generation unit 22 are input (each multivalued image data input step, SS150), each of the multivalued image data 28 is preprocessed with some kind of texture. It is determined whether image processing that handles information (for example, image processing that emphasizes texture) has been performed (preliminary texture information processing detection step, SS152), and each multi-valued image data 28 is preliminarily subjected to any texture information. If no processing (for example, image processing for emphasizing texture) has been performed, the texture information of each multivalued image data 28 is detected by the texture detection unit 30 (texture detection step, SS154).

Here, texture information means information on texture portions. Physically, the texture part is a place where the luminance gradient in the image is high and the luminance gradient dispersion in the local area is high. It refers to parts such as contours (edges). The texture information includes, for example, surface unevenness information, surface pattern information, brightness information, and contour information or edge information in a collection of fine objects of such a texture portion. These pieces of information are perceptually referred to as 'unevenness', 'pattern', 'edge', etc., depending on the difference in values between the luminance gradient after noise removal and the local luminance gradient variance information.

In the texture detection step SS154, for example, when the texture information is edge information, noise components are separated from the multilevel input image data by noise removal in advance, and then luminance signals (brightness) are extracted from the multilevel input image data. The texture information is detected by extracting the edge intensity using the gradient and calculating the texture intensity using the local area variance of the extracted edge intensity. Note that if the noise is not removed, the noise component is also regarded as a texture portion, so it is preferable to separate it in advance.

Further, when each of the multi-valued image data 28 has been previously subjected to image processing that handles some texture information (for example, image processing that emphasizes texture), the texture information of each of the multi-valued image data 28 is is used (step of using texture information, SS156). In this case, since existing texture information is used, there is an advantage that the amount of calculation for detecting texture information is reduced.

The error distribution unit 32 has a function of determining the distribution of the quantization error of the error diffusion method according to the strength of the texture information. is realized (error distribution step, SS158).

An error diffusion unit 34 converts each multivalued image data 28 into binary quantized image data 36 by executing gradation conversion processing on each multivalued image data 28 so as to emphasize the texture information. (quantization step in the error diffusion unit, S160). The error diffusion section 34 performs quantization based on the error distribution result determined by the error distribution section 32 . Then, all pixels are repeatedly processed in order by raster scanning, vector scanning, or the like pixel by pixel (step of determining whether or not all pixels have been quantized, SS162).

After all pixels are quantized, they are output as binary quantized image data 36 (binary quantized image data output step, SS164). If there are remaining pixels, the process returns to the error distribution step SS158 in the error distribution unit 32 and repeats.

In the example of the quantization step S106a by the image processing device 10A described above, an example in which the prior texture information processing detection step SS152 is provided has been shown, but another example of the quantization step S106a by the image processing device 10A is 2) Without providing the preliminary texture information processing detection step SS152, each of the multivalued image data 28 is processed in the case where the preliminary texture information processing is performed and in the case where the preliminary texture information processing is not performed. can also

By inputting the two types of binary quantized image data generated in this way (binary quantized image data input step, S108), the multi-valued quantized image generation unit 40 outputs four final images. A value quantized image is generated (multi-value quantized image generating step, S110). In this way, the multilevel quantized output image data 42 of FIG. 1 is created and finally output as multilevel quantized output image data as shown in FIG. 2 (output step, S112).

Quantization to four values (a<b<c<d) in each pixel in the multi-value quantized image generation unit 40 is realized by the following algorithm.
a, if Pixel value of multilevel input image = 0
b, if pixel value of binary quantized image 1 = 0 and pixel value of binary quantized image 2 = 0
c, if pixel value of binary quantized image 1 = 0 and pixel value of binary quantized image 2 = 1
Or, pixel value of binary quantized image 1 = 1 and binary quantized image 2
pixel value=0
d, if pixel value of binary quantized image 1 = 1 and pixel value of binary quantized image 2 = 1
Here, the values of a, b, c, and d are, for example, a = 0 (black), b = 1/3 (dark gray), c = 2/3 (light gray), d = 1 (white). Good luck. As a result, black in the multilevel input image data can be represented as black even after the multilevel quantized output image data. In addition, it is possible to obtain an output result with good dot dispersion by performing a logical operation of two types of binary quantized images with different lightness for halftones.

As long as the number of gradations is smaller than the number of input gradations, the present invention is not limited to quaternary quantization, and can be applied to quantization of any number of gradations.

In the above description, the level image generating unit 22 in each multivalued image generating step S102 shows an example of conversion to 2 bits (quaternary) with N=2, but conversion to other values is also possible. be. For example, when converting to N=3 bits (eight values), for example,
Y= ^xn , Y=x, Y=x1 ^/n , 0≤n≤1
This can be achieved by setting three types of conversion functions. It is important to perform nonlinear transformation to multivalued images with different average brightness.

Furthermore, another algorithm for the multilevel quantized image generation step S110 in the multilevel quantized image generation unit 40, that is, an example of a conversion rule is as follows.
a, if pixel value of binary quantized image 1 = 0 and pixel value of binary quantized image 2 = 0
b, if pixel value of binary quantized image 1 = 0 and pixel value of binary quantized image 2 = 1
Or pixel value of binary quantized image 1 = 1 and pixel value of binary quantized image 2 = 0
c, if pixel value of binary quantized image 1 = 1 and pixel value of binary quantized image 2 = 1
d, if Pixel value of multilevel input image = 1
It is also possible to set the conversion rule as

<Image processing device 10B>
A schematic configuration of the image processing apparatus 10B is shown in the block diagram of FIG. As shown in FIG. 4, the image processing apparatus 10B is an image processing apparatus that converts multi-valued input image data 50 into multi-valued quantized output image data 90. For the multi-valued input image data 50, brightness a level image generator 52 for generating a plurality of multi-valued image data 62 (multi-valued image data level 1 to multi-valued image data level N) having different values, and texture detection for detecting texture information of each multi-valued image data 62 A unit 64 converts each multi-valued image data 62 into each binary quantized image data 76 (binary image data 76) by executing gradation conversion processing on each multi-valued image data 62 so as to emphasize the texture information. an error diffusion unit 74 for converting data 1 to binary image data N), and a multi-value quantized image generation unit 80 for generating multi-value quantized output image data based on the plurality of binary quantized image data 76. and an image processing apparatus.

The image processing apparatus 10B further includes a difference calculation unit 68 for generating a difference image of each of the multilevel input image data 62 and each of the binary quantized image data quantized by the error diffusion unit 74; A visual processing unit 70 that is a visual characteristic filter to be input, a correction amount calculation unit 66 that calculates a correction amount so as to preserve the texture information, and a correction amount for each of the multivalued input image data 62 based on the correction amount. and a multivalued image correction unit 72 , and the error diffusion unit 74 generates the binary quantized image data 76 based on the output of the multivalued image correction unit 72 .

In the present invention, multilevel input image data 50 can be converted into N-bit multilevel quantized output image data 90 . Therefore, in the image processing apparatus 10B as well, each of the plurality of multi-valued image data 62 has any number from multi-valued image data level 1 to multi-valued image data level N according to the number of bits of the output multi-valued quantized image. Level. Then, according to the number of each multivalued image data 62 (level 1 to level N), the quantization unit 60 converts each binary quantized image data 76 (binary image data 1 to binary image data N) into will be created. The schematic configuration of the image processing device 10B is basically the same as that of the image processing device 10A, and only the configuration of the quantization section 60 differs from that of the quantization section 26 of the image processing device 10A.

The image processing apparatus 10B, like the image processing apparatus 10A described above, is a computer including a control unit such as a CPU and a storage unit, and is caused to function as each means by a program so as to perform each process of the image processing apparatus 10A. . Further, each process of the image processing apparatus 10A may be performed via a network such as the Internet.

The multi-valued input image data 50 is the same as the multi-valued input image data 20, and is a color or monochrome digital image expressed in 8-bit gradation or the like. multi-value quantized output image data 90 is obtained.

One embodiment of the image processing method by the image processing device 10B will be described with reference to FIGS. 2 and 4 to 5. FIG. The image processing method can be executed by a program stored in a memory of a computer.

The image processing method of the present invention by the image processing apparatus 10B is an image processing method for converting monochrome or color multi-valued input image data 50 into monochrome or color multi-valued quantized output image data 90. FIG. As an example, a case of outputting a four-level quantized image (2 bits: N=2) as the multi-level quantized output image data 90 will be described below with reference to FIGS. 2, 4 and 5. FIG.

First, monochrome or color multi-valued input image data 50 is input from another information terminal or storage medium (multi-valued image data input step, S100).

As with the image processing apparatus 10A described above, the level image generation unit 52 generates two types of multivalued images from the multivalued input image data 50 (multivalued image generation step, S102). The quantization unit 60 applies two types of binary quantization to each multivalued image data 62 (multivalued image data level 1 and multivalued image data level 2), which are two types of generated images with different brightness. A quantized image is generated (quantization step, S106b).

Sub-steps in the quantization step S106b will be described with reference to FIG.

When the two types of multivalued image data 62 generated by the level image generation unit 52 are input (each multivalued image data input step, SS200), each of the multivalued image data 62 is preprocessed with some kind of texture. It is determined whether image processing that handles information (for example, image processing that emphasizes texture) has been performed (preliminary texture information processing detection step, SS202), and each multivalued image data 62 is preliminarily subjected to any texture information. (for example, image processing that emphasizes the texture) is not performed, the texture information of each of the multi-valued image data 62 is detected by the texture detection unit 64 (texture detection step, SS204). . The texture information and the texture detection step SS204 are as described in the above image processing apparatus 10A.

Further, when each of the multi-valued image data 62 has been previously subjected to image processing that handles some texture information (for example, image processing that emphasizes texture), the texture information of each of the multi-valued image data 62 is Use

Further, each of the multi-valued image data 62, which has been previously subjected to image processing that handles some texture information, is corrected by the multi-valued image correction unit 72 (multi-valued image correction step, SS212), and error diffusion is performed. It is quantized in unit 74 (quantization step, SS214).

The binary quantized image data quantized by the error diffusion unit 74 in SS214 is sent to the difference calculation unit 68, and the multivalued image data 62 and the binary quantized image data quantized in SS214 are is generated by the difference calculation unit 68 (difference image generation step, SS206).

This generated difference image is passed through a visual characteristic filter by the visual processing unit 70 (filter processing step, SS208), and then processed by the correction amount calculation unit 66 to obtain the texture obtained by the texture detection unit 64 in the texture detection step SS204. information, and the correction amount calculation section 66 derives the correction amount so that the multi-value quantized image preserves the texture information (correction amount deriving step, SS210).

Then, based on the correction amount in the correction amount derivation step SS210 by the correction amount calculation unit 66, the input multivalued input image is corrected by the multivalued image correction unit 72 (multivalued image correction step, SS212). It is quantized by the error diffusion unit 74 (quantization step, SS214). Then, all pixels are repeatedly processed in order by raster scanning, vector scanning, or the like pixel by pixel (step of determining whether or not all pixels have been quantized, SS216).

After all pixels are quantized, they are output as binary quantized image data 76 (binary quantized image data output step, SS218). If there are remaining pixels, the process returns to the difference image generation step SS206 in the difference calculation unit 68 and repeats.

In the example of the quantization step S106b by the image processing device 10B described above, an example in which the preliminary texture information processing detection step SS202 is provided is shown, but another example of the quantization step S106b by the image processing device 10B is 3. Without providing the preliminary texture information processing detection step SS202, each of the multi-valued image data 62 is processed in the case where the preliminary texture information processing is performed and in the case where the preliminary texture information processing is not performed. can also

By inputting the two kinds of binary quantized image data generated in this way (binary quantized image data input step, S108 in FIG. 2), the multi-valued quantized image generation unit 80 finally outputs A four-level quantized image is generated (multi-level quantized image generating step, S110 in FIG. 2). In this way, the multilevel quantized output image data 90 of FIG. 4 is created and finally output as multilevel quantized output image data as shown in FIG. 2 (output step, S112).

The algorithm for quantizing each pixel into four values or the like in the multi-value quantized image generation unit 40 is the same as that of the image processing apparatus 10A described above.

By iteratively repeating the above process, perceptually texture-preserving quantization can be achieved. The image processing method of the present invention by the image processing apparatus 10B has the advantage of generally obtaining good quantization results compared to the image processing method by the image processing apparatus 10A.

The error diffusion units 34 and 74 perform quantization on all pixels in order by raster scanning, vector scanning, or the like, pixel by pixel, using an error diffusion method. The error diffusion method is a type of method that expresses images smoothly by halftone processing. This is a method of minimizing the error as a whole by allocating the error generated in the processing of a pixel of a digital image to surrounding pixels, and then performing processing in consideration of the effect of allocating the error. As an example, a case of outputting a quaternary quantized image with inks of C (cyan), M (magenta), Y (yellow), and K (black) will be described.

For example, when inks of C, M, Y, and K are used, CMYK each has four values, and four values are determined for each ink by the following method.
<For C (cyan)>
output = 0 (dark cyan), if X ≤ a ₁ ,
output = 85 (middle cyan), if a ₁ < X ≤ a ₂ ,
output = 170 (light cyan), if a ₂ < X ≤ a ₃ ,
output = 255 (no ink placed), if a ₃ < X
where a ₁ , a ₂ , and a ₃ are threshold parameters determined by droplet size and ink density, determined by current color proofing techniques to minimize the color difference between input and printed image. For each of the other M, Y, and K inks, four values are determined in the same manner as described above.

As long as the number of gradations is smaller than the number of input gradations, the present invention is not limited to quaternary quantization, and can be applied to quantization of any number of gradations. In addition, since various factors in the printing process, such as ink characteristics, paper characteristics, and head characteristics, accumulate to cause non-linearity, for example, a color chart of a digital image is printed, and the color of the image and the print result are measured. Four values are determined by estimating The error diffusion units 34 and 74 diffuse the error based on this threshold parameter, and quantize each pixel in order by raster scanning, vector scanning, or the like. M, Y, and K inks other than C are also quantized in the same manner.

As another example, a case of outputting a hexanary quantized image with inks of C (cyan), M (magenta), Y (yellow), and K (black) will be described. For example, when inks of C, M, Y, and K are used, CMYK each have hexavalues, and the hexavalues are determined for each ink by the following method.
<For C (cyan)>
output = 0 (dark cyan), if X ≤ a ₁ ,
output = 51 (slightly dark cyan), if a ₁ < X ≤ a ₂ ,
output = 102 (middle cyan), if a ₂ < X ≤ a ₃ ,
output = 153 (light cyan), if a ₃ < X ≤ a ₄ ,
output = 204 (light cyan), if a ₄ < X ≤ a ₅ ,
output = 255 (no ink placed), if a ₅ < X
Here, a ₁ , a ₂ , a ₃ , a ₄ , and a ₅ are threshold parameters determined by droplet size and ink density, as described above. The error diffusion units 34 and 74 diffuse errors based on these threshold parameters, and quantize each pixel in order by raster scanning, vector scanning, or the like. Six values are determined in the above-described manner for each of M, Y, and K inks other than C, and quantization is performed in the same manner.

FIG. 7 shows an example of processing steps in each stage. FIG. 7 is an image of the cyan component of the color image. FIG. 7(a) shows a multivalued input image (8 bits), and FIGS. 7(b) and 7(c) show two kinds of multivalued images (that is, each multivalued image data). 7(d) and 7(e) are the results of binary quantization by the quantization section 26 of the image processing apparatus 10A described with reference to FIGS. 1 to 3. FIG. FIG. 7(f) shows the multilevel quantized output image data 42 generated by the multilevel quantized image generation unit 40, which is the quantization result according to the present invention.

FIG. 8 shows the results of printing photographic images according to Example 1 and Comparative Example 1 of the present invention. In Example 1 and Comparative Example 1, an inkjet printer FXIJ manufactured by Think Laboratory Co., Ltd. was used, and printing was performed using a 600 dpi inkjet KJ48 for water-based inks and using water-based inks of C, M, Y, and K. . For each droplet size, 0 pL (ink discharge amount zero picoliter, no dots), 3 pL, 4 pL, and 5 pL were used.

<Comparative Example 1>
FIG. 8A shows the results of inkjet printing of image data obtained by quantizing the conventional error diffusion method into four values at 600 dpi. In the result of the conventional method shown in FIG. 8A, the texture is not sufficiently reproduced because the error is spatially diffused without considering the structure of the texture. Further into the printing process, the ink almost crushed the texture.

<Example 1>
FIG. 8(b) is the result of inkjet printing multi-level quantized output image data obtained by four-level quantization at 600 dpi using the image processing apparatus 10A and the image processing method of the present invention shown in FIGS. is. Four values were determined for each ink in the same manner as in the case of using the C, M, Y, and K inks described above. The error diffusion unit 34 diffuses the error based on this threshold parameter, and quantizes each pixel in order by raster scanning, vector scanning, or the like. Quantization was performed in the same manner for each ink. In FIG. 8B according to the present invention, the texture of the input image is quantized and expressed in a form that emphasizes it so that it can be expressed in texture such as fine fur and surface reflection. Therefore, even after the printing process, it is possible to obtain a print result in which texture information is preserved.

<Comparative Example 2 and Example 2>
Further, as an embodiment performed by a method similar to the above, in the case of quaternarization by the conventional error diffusion method, if a color that is spatially uniform is input, it is substantially binarized. For example, assuming that (0, 85, 170, 255) inks are used when densities are replaced by pixel values, only 85 and 170 colors are used for 128 pixel value colors. be. Therefore, in the conventional error diffusion method, artifacts in which a specific pattern appears as shown in FIG. 9A are conspicuous (comparative example 2). On the other hand, according to the present invention, since the colors are converted into two kinds of light and dark colors, and synthesized after binarizing each color, it is possible to substantially realize multi-value conversion, as shown in FIG. 9(b). Obtrusive artifacts can also be reduced (Embodiment 2).

As described above, according to the present invention, compared with the printing result of the conventional error diffusion method, the texture and the The superiority that can improve the texture has been confirmed.

Furthermore, the conventional error diffusion method is also used as a technique for expressing low luminance contrast in high luminance displays. Also in the technical field of displays, enhancing textures in images is an important issue. Therefore, if the quantization technique of the present invention is applied to a display, a display such as a television that can be applied to emphasize texture can be realized.

In addition, although the specific configuration of the texture detection unit is not shown in the above description, for example, the intensity of the edge of the image is extracted by a differential filter. Since the extracted high-intensity edges contain noise in the image and object contours, it is necessary to remove them from the extracted edges in order to detect the texture information. It is possible to detect only texture information by assuming that a large density variance value in a local region of each pixel is noise or the outline of an object, and multiplying the edge intensity by the reciprocal thereof.

As texture information, vertical lines, transparency (brightness information), edge information, or contour information have been described, but other texture information such as texture of an object, unevenness information related to surface conditions, etc. may be used.

Moreover, frequency subband information and the like can be considered as the existing texture information used for texture enhancement processing, but other texture information may be used.

In addition, in the image processing method (FIGS. 2 and 5) using the image processing device 10B, the process of determining the amount of correction and performing control based on the amount of correction was iteratively repeated. The larger the number of iterations, the better the quantization result can be obtained. Moreover, the structure which does not repeat repetitively may be sufficient, and in that case, the amount of calculations can be minimized.

INDUSTRIAL APPLICABILITY The present invention is industrially applicable as an image processing apparatus, an image processing method, and a program that can be applied to enhance textures in images such as photographs. The present invention can be suitably used for commercial printers, home printers, displays, televisions, digital cameras, image processing software, and the like.

10A, 10B: image processing device, 12a, 12b: roll, 14: web-like printing substrate, 16a, 16b, 16c, 16d: inkjet head, 18: head controller, 20, 50: multi-value input image data, 22 , 52: level image generation unit, 26, 60: quantization unit, 28, 62: each multivalued image data, 30, 64: texture detection unit, 32: error distribution unit, 34, 74: error diffusion unit, 36, 76: Binary quantized image data 40, 80: Multilevel quantized image generator 42, 90: Multilevel quantized output image data 66: Correction amount calculation unit 68: Difference calculation unit 70: Visual processing Section 72: Multi-value image correction section 100: Inkjet printer.

Claims (10)

An image processing device for converting multilevel input image data into multilevel quantized output image data,
a level image generation unit that generates a plurality of multivalued image data with different brightness for the multivalued input image data;
a texture detection unit that detects texture information of each of the multivalued image data;
an error diffusion unit that converts each of the multi-valued image data into each of binary quantized image data by executing tone conversion processing on each of the multi-valued image data so as to emphasize the texture information;
a multi-level quantized image generating unit that generates multi-level quantized output image data based on the plurality of binary quantized image data;
An image processing device having 2. The image processing apparatus according to claim 1, wherein said texture information is surface unevenness information, surface pattern information, brightness information, edge information or contour information. 3. The image processing apparatus according to claim 1, further comprising an error distribution unit that calculates a distribution of quantization error of said transform according to said texture information. 4. The image processing apparatus according to claim 1, wherein said texture information is texture enhancement processing information of said multivalued input image data. 5. The image processing apparatus according to claim 4, wherein the texture enhancement processing information is frequency subband information. a difference calculation unit that generates a difference image of each of the multilevel input image data and each of the binary quantized image data quantized by the error diffusion unit; and a visual processing unit that is a visual characteristic filter that inputs the difference image. , a correction amount calculation unit that calculates a correction amount so as to preserve the texture information, and a multi-valued image correction unit that corrects each of the multi-valued input image data based on the correction amount,
6. The image processing apparatus according to any one of claims 1 to 5, wherein the error diffusion unit generates each of the binary quantized image data based on the output of the multivalued image correction unit. An image processing method for converting multilevel input image data into multilevel quantized output image data, comprising:
each multi-valued image generating step of generating a plurality of multi-valued image data with different brightness in a level image generation unit for the multi-valued input image data;
a texture detection step of detecting texture information of each of the multivalued image data by a texture detection unit;
a step of converting each multi-valued image data into each binary quantized image data in an error diffusion unit by executing a gradation conversion process on each multi-valued image data so as to emphasize the texture information;
a step of generating multilevel quantized output image data in a multilevel quantized image generation unit based on each of the plurality of binary quantized image data;
An image processing method comprising: a step of generating a differential image of the binary quantized image data quantized by the error diffusion unit and each of the multi-valued input image data by a difference calculation unit; a step of calculating a correction amount by a correction amount calculation unit so as to store the texture information; and a step of correcting each of the multivalued input image data by the multivalued image correction unit based on the correction amount. 8. The image processing method according to claim 7, further comprising: generating the binary quantized image data in the error diffusion section based on the output of the multi-valued image correction section. 9. The image processing method according to claim 8, wherein the step of calculating the correction amount by said correction amount calculation unit is repeated. A program for causing a computer to function as each means of the image processing apparatus according to any one of claims 1 to 6.

Images